Systems and methods for providing a systemic error in artificial intelligence algorithms
a technology of artificial intelligence and system error, applied in the field of training neural networks and artificial intelligence or machine learning algorithms or models, can solve the problem that adversity can unfairly be the equivalent of taking the original model withou
Active Publication Date: 2022-01-27
TRIPLEBLIND INC
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Benefits of technology
The patent text describes a method for training neural networks and detecting if a competitor has used a different model trained on the same algorithm. This is important because it helps to protect the company's intellectual property. The method involves creating a set of fingerprints and testing if a suspect model was derived from a source model. This approach helps to ensure that the training data is used appropriately and prevents unfair use of the company's model. The patent also describes methods for adding noise to a sub-group of adversarial candidates, requesting a marking key from a model owner, and applying a fingerprint to a suspect model to determine if it was derived from a source model. The technical effects of this patent are improved training of neural networks and better protection of company's intellectual property.
Problems solved by technology
Such an approach by the adversary can unfairly be the equivalent of taking the original model without compensation.
Method used
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0033]Disclosed herein is a new system, a platform, compute environment, cloud environment, marketplace, or any other characterization of the system that will enable an improved approach to training neural networks. In one aspect, the approach is called a federated-split leaning approach that combines features from known approaches but that provides a training process that maintains privacy for data used to train the model from various client devices. Thus, entities that have models that they desire to be protected can provide a dataset that provides some or all of the data used to train their source model for use in developing the shadow models. In one aspect, only a computer system receives the data and generates the shadow models. In other words, in one aspect, no human is able to access the received dataset used to develop the shadow models.
[0034]The general concept disclosed herein is illustrated by an example method. An example method includes receiving, from a model owner nod...
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More PUM
Login to View More Abstract
Disclosed is a process for testing a suspect model to determine whether it was derived from a source model. An example method includes receiving, from a model owner node, a source model and a fingerprint associated with the source model, receiving a suspect model at a service node, based on a request to test the suspect model, applying the fingerprint to the suspect model to generate an output and, when the output has an accuracy that is equal to or greater than a threshold, determining that the suspect model is derived from the source model. Imperceptible noise can be used to generate the fingerprint which can cause predictable outputs from the source model and a potential derivative thereof.
Description
PRIORITY CLAIM[0001]This application is a non-provisional patent application of U.S. Provisional Application No. 63 / 090,933, filed on Oct. 13, 2020, which is incorporated herein by reference.[0002]This application is a continuation-in-part of U.S. patent application Ser. No. 16 / 828,085, filed Mar. 24, 2020, which claims the benefit of U.S. Provisional Application No. 62 / 948,105, filed Dec. 13, 2019, which is incorporated herein by reference.[0003]This application is a continuation-in-part of U.S. patent application Ser. No. 16 / 828,216, filed Mar. 24, 2020, which claims the benefit of U.S. Provisional Application No. 62 / 948,105, filed Dec. 13, 2019, which is incorporated herein by reference.[0004]This application is a continuation-in-part of U.S. patent application Ser. No. 17 / 176,530, filed Feb. 16, 2021, which is a continuation of U.S. patent application Ser. No. 16 / 828,354, filed Mar. 24, 2020, now U.S. Pat. No. 10,924,460, issued on Feb. 16, 2021, which claims the benefit of U.S....
Claims
the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More Application Information
Patent Timeline
Login to View More IPC IPC(8): H04L29/06G06N3/04H04L9/00G06Q20/40G06K9/62G06F17/16G06Q30/06H04L9/06G06N3/08
CPCH04L63/0428G06N3/04H04L9/008G06Q20/401G06K9/6267G06Q2220/00G06F17/16G06Q30/0623H04L9/0625G06N3/082H04L2209/46G06K9/623H04L9/3231H04L9/3239G06N3/08G06V10/764G06V10/774G06N3/045G06F18/24G06F18/2113
Inventor GHARIBI, GHARIBGILKALAYE, BABAK POOREBRAHIMDAS, RIDDHIMAN
Owner TRIPLEBLIND INC